Skip to main content

  1. Agentycs
  2. Deployment
  3. Private cloud

Private cloud

Your cloud account, your Kubernetes, our platform.


What it is #

In a private cloud deployment the platform is a tenant of your infrastructure rather than the other way round. You bring a Kubernetes cluster and its storage classes; the Agentycs operator reconciles the rest — the platform foundation, the database tier, object storage, the inference plane and the applications — from custom resources held in your own git repository.

Because the whole platform is declared rather than assembled, the environment you validated is the environment that runs. Upgrades are a change to a declared version, and the operator converges the cluster to it. Nothing depends on a technician remembering the order of the steps.

Who it is for

  • You already run production workloads in a cloud account and intend to keep them there.
  • Your network policy is yours, and egress must stay under it.
  • You want the platform's capabilities without moving data to another operator.
  • Your platform team is comfortable operating Kubernetes and wants declarative control.

The environment you validated is the environment that runs, because it was declared rather than assembled.

What you get #

The whole platform runs here. These are the parts this pattern changes the shape of.

01

Declared, not installed #

One operator reconciles the platform instance, tenants, apps, databases and each subsystem cluster from custom resources, and keeps them converged after every change.

02

Your storage, or ours #

Datasets rest on platform object storage in-cluster, or on your own S3-compatible backend. Where data lives is a per-tenant choice, not a platform assumption.

03

Egress on your terms #

Identity, artefact, telemetry and support paths are explicit, so your proxy and firewall policy can allow exactly what the platform needs and nothing else.

04

Your accelerators, whatever they are #

The inference plane is accelerator-agnostic: models are placed only on hardware whose architecture supports their kernels and numeric format, across mixed GPU generations.

The operating model #

Nobody should discover who owns a failure while it is happening. This split is explicit before anything is signed, and it is the part of a deployment decision that outlives the architecture.

You operate
The cloud account, the cluster, its capacity and its network policy.
The operator handles
Platform reconciliation, tenant composition, database and storage lifecycle, model placement and rollout.
We can operate alongside
Platform on-call, upgrade shepherding and diagnostics, under a support agreement and with access you grant explicitly.
You keep
Identity, keys, tenant policy, audit export and every decision about what may leave the account.